Blog
Playbook · · 7 min read
Where to draw the line between your AI agent and your team
Draw the line too tight and the agent hands over questions it could have answered. Draw it too loose and it keeps talking to a customer who asked for a person three messages ago. Here's how to set it, and who should own it.
Playbook ·
How to tell whether your AI agent is actually working
The five numbers on an AI support agent's analytics page, the one to watch first, how to set a target you didn't borrow, and how to turn a metric into a document fix.
Playbook ·
Every price change is a release: a checklist for keeping your AI agent right
A grounded AI agent quotes whatever your documentation said last. A three-step routine for keeping it current when a price, a delivery zone, or a holiday closing time changes.
Playbook ·
How to make your AI agent sound like your team, not like a model
Conversation design for a small support team: the three voice settings, why your documents are the script, how to word a handoff, and where to start without hiring anyone.
Playbook ·
How many seats does a small support team actually need?
Capacity planning with an AI agent, in Anchor's credit model: count your conversations, size the pool, plan for the humans' work getting heavier, and know what happens at zero.
Playbook ·
Rolling out an AI agent to a support team of two
Change management for a small support team adopting an AI agent: what to say before go-live, how the day changes, which numbers to share, and who owns the improvement work.
Field notes ·
What to ask an AI support vendor about your data when you don't have a security team
A short list of questions to ask any AI support vendor about customer data — the ordinary six, the two that are new with AI agents, and how Anchor answers each, including the one that's 'not yet'.
Scorecard ·
Six things an AI support agent does better than a human
Where an AI support agent beats a human agent — speed, volume, hours, patience, languages — where it still loses, and the split that gets you both.